Accessibility & Standards

Tagged PDF

Definition

Strip a PDF back and you find a page description: glyphs placed at coordinates, with no idea which of them form a sentence. Tagging adds a second description of the same content as a hierarchy, stored in the file under an object called StructTreeRoot. The element names are borrowed from HTML and read much the same: H1 through H6, P, L and LI for lists, Table with TR, TH and TD, Figure for an image, Link for a link. Page furniture that carries no meaning — a rule under the running header, a decorative border — is marked as an artifact instead, so it is skipped rather than announced. Tagging arrived with PDF 1.4 in 2001 and is required by PDF/UA, the accessibility standard published as ISO 14289, which also requires alt text on every Figure. A file can be tagged badly, so the presence of a tag tree is a starting point, not a verdict.

Why It Matters

Tags are the difference between an accessible PDF and a picture of one. With them, a reader can jump heading to heading, hear a table announced cell by cell with its column header, and reflow the text to a phone screen at large type; without them, the document is one long undifferentiated run, or silence. Copying text out of an untagged two-column layout usually interleaves the columns, which is the same ordering problem showing up in a way sighted readers can see. Conversion also depends on tags: exporting a PDF to HTML, EPUB or Word produces something usable only when the structure was there to start with.

How It Works

In Acrobat Pro the tree appears in the Tags panel, where each element can be renamed, nested or deleted, and the Order panel shows the sequence in which content is read. Auto-tagging guesses from visual layout, so it tends to get body text right and tables, sidebars and multi-column pages wrong, which is why the reading order needs checking by eye after any automatic pass. Tables want real TH cells with a Scope of Column or Row, figures want an Alt entry, and anything purely decorative wants marking as an artifact so it disappears from the read-out. Heading levels must nest without gaps, the same rule that applies in HTML. The cheap route is to skip all of this by exporting correctly from Word, InDesign or LibreOffice in the first place, since those tools already know which paragraph was a heading.

Real-World Example

A consultancy publishes a 2 MB case study at client-story.99helpers.site with a two-column layout. Untagged, the text read across both columns and came out as alternating half-sentences. After re-exporting from InDesign with the articles panel set and tagging enabled, the same file read down column one, then column two, with the pull quote tagged as a figure and skipped in the main flow. The file on 99helpers was replaced in place, so the shared link never changed and the previous version stayed available in version history.

Common Mistakes

  • Running auto-tag on a designed layout and shipping the result — it reads visual position, not intent, and reliably mangles columns, sidebars and tables
  • Marking a data table's header row as ordinary cells — a screen reader then reads bare numbers with no column name attached, which makes the table useless to listen to
  • Leaving decorative rules, page numbers and background art untagged as content — every one of them gets announced, and the reader hears noise between paragraphs

Related Terms

Put a file online in seconds

Drop in a document, an image, a page or a whole static website and share the link — free, with no build step and no server to set up.

Host a file free →